Sample Size Issues in the Design of Clinical Trials
Monday 2 June, 10:20--12:10

Organizer: Jackie HAN
Bristol-Myers Squibb
Moderator: Jackie HAN
Bristol-Myers Squibb

Power and Sample Size for Multiple Comparisons of Means of Normal Distributions using SAS

David R. BRISTOL
Schering-Plough Research Institute, Kenilworth, NJ, USA
Until recently, hypothesis testing using multiple comparison procedures (MCPs) has required extensive tables of critical values. The function PROBMC of SAS (1992) provides a solution to the problem of providing p-values for various MCPs; namely, Dunnett's procedure for comparison with a control, the Studentized Range Test, and the Studentized Maximum Modulus

Test. The problem of computing the power of these test procedures and the corresponding problem of sample size determination to control the power have not been adequately addressed. In some cases, PROBMC can be used to evaluate the power; some simple numerical methods are presented for use in other situations. Some simulation results are presented to examine the adequacy of the approximations.

[David R. Bristol, Schering-Plough Research Institute, Kenilworth, NJ 07033]


Re-estimation of Sample Size for Triple Blind Clinical Trials

Weichung Joe SHIH
Merck Research Laboratories, Rahway, NJ, USA
Sample size re-estimation for clinical trials is a type of interim analysis with a purpose differing from that of the familiar group sequential methods. In this talk I will introduce the rationale, methods, and issues for two common cases: normal and binary outcome variables. Current regulatory guidelines for statistical design and analysis regarding the sample size re-estimation in new drug applications will also be discussed.

[Weichung Joe Shih, Merck Research Laboratories]


Sample Size Calculation for Clinical Trials using Commercial Software

Hongjie Zheng, Ph.D
Novartis Pharmaceuticals, Summit, NJ, USA
Sample size calculation for clinical trials is routine for statisticians working in the pharmaceutical industry. Often statisticians calculate sample size either by hand or by writing their own little computer programs. Since sample size calculation is so important in determining the cost of the trial as well as in demonstrating the efficacy of a new drug when the new drug is indeed efficacious, a well validated sample size software will be a useful tool to aid the statisticians in their calculation. In this talk, I will review the top two ranked commercial software PASS 6.0 and nQuery Advisor in terms of their capacity for sample size calculation in clinical trials. Both statistical procedures and their accuracy of the packages considered will be examined, but focus will be on the usefulness of the software in clinical trials.

[Hongjie Zheng, Ph.D, Department of Biostatistics, Novartis Pharmaceuticals, Summit, NJ 07901]


Back to JRC program
Back to JRC Home Page